Delft University of Technology

An extensive numerical benchmark of the various magnetohydrodynamic flows

Blishchik, Artem; van der Lans, Mike; Kenjereš, Saša

DOI
10.1016/j.ijheatfluidflow.2021.108800
Publication date
2021
Document Version
Final published version
Published in
International Journal of Heat and Fluid Flow

Citation (APA)
Blishchik, A., van der Lans, M., & Kenjereš, S. (2021). An extensive numerical benchmark of the various
magnetohydrodynamic flows. International Journal of Heat and Fluid Flow, 90, [108800].
https://doi.org/10.1016/j.ijheatfluidflow.2021.108800

Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.

Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

For technical reasons the number of authors shown on this cover page is limited to a maximum of 10.
 International Journal of Heat and Fluid Flow 90 (2021) 108800

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier.com/locate/ijhff

An extensive numerical benchmark of the various

magnetohydrodynamic flows
Artem Blishchik, Mike van der Lans, Saša Kenjereš *
Transport Phenomena Section, Department of Chemical Engineering, Faculty of Applied Sciences, Delft University of Technology and J.M. Burgerscentrum Research
School for Fluid Mechanics, Van der Maasweg 9, 2629 HZ Delft, The Netherlands

A R T I C L E I N F O A B S T R A C T

Keywords: There is a continuous need for an updated series of numerical benchmarks dealing with various aspects of the
MHD magnetohydrodynamics (MHD) phenomena (i.e. interactions of the flow of an electrically conducting fluid and
Magnetic field an externally imposed magnetic field). The focus of the present study is numerical magnetohydrodynamics
Lorentz force
(MHD) where we have performed an extensive series of simulations for generic configurations, including: (i) a
OpenFOAM
Numerical benchmark
laminar conjugate MHD flow in a duct with varied electrical conductivity of the walls, (ii) a back-step flow, (iii) a
multiphase cavity flow, (iv) a rising bubble in liquid metal and (v) a turbulent conjugate MHD flow in a duct with
varied electrical conductivity of surrounding walls. All considered benchmark situations are for the one-way
coupled MHD approach, where the induced magnetic field is negligible. The governing equations describing
the one-way coupled MHD phenomena are numerically implemented in the open-source code OpenFOAM. The
novel elements of the numerical algorithm include fully-conservative forms of the discretized Lorentz force in the
momentum equation and divergence-free current density, the conjugate MHD (coupling of the wall/fluid do­
mains), the multi-phase MHD, and, finally, the MHD turbulence. The multi-phase phenomena are simulated with
the Volume of Fluid (VOF) approach, whereas the MHD turbulence is simulated with the dynamic Large-Eddy
Simulation (LES) method. For all considered benchmark cases, a very good agreement is obtained with avail­
able analytical solutions and other numerical results in the literature. The presented extensive numerical
benchmarks are expected to be potentially useful for developers of the numerical codes used to simulate various
types of the complex MHD phenomena.

1. Introduction
One of the pre-requisites to be able to deal with advanced physical
transport phenomena involving the magnetohydrodynamics (MHD) in­
teractions is to have a well-validated and numerically efficient computer
code. This still poses a quite challenging task due to a lack of advanced
experimental studies that can provide detailed insights into the flow and
electromagnetic parameters that can be used to validate computer
codes. The essence of the MHD phenomena is usually associated with a
flow of highly electrically conducting liquid metals, which are, due to
their non-transparency, notoriously difficult to study with standard
laser-based optics diagnostics tools.
To validate MHD numerical models, we have to rely on analytical
solutions that are based on significant simplifications. In the present
manuscript, we are revisiting and proposing an extensive list of possible
benchmark cases available in the open literature dealing with various
aspects of the MHD phenomena. One of the simplest numerical MHD
benchmarks is a fully developed laminar channel, duct, or pipe flow
subjected to a uniform magnetic field of different orientations, for which
an exact analytical solution exists, Hartmann and Lazarus (1937),
Shercliff (1953). The effects of the non-uniform longitudinal magnetic
field on a laminar flow of electrically conducting fluid in a pipe were
recently numerically simulated in Feng et al. (2015). The open-source
computer code OpenFOAM was used and good agreement was ob­
tained between simulations and experiments. The MHD flow in a duct
with very thin electrically conducting walls was presented in Tao and Ni
(2013). Instead of fully resolving the wall region, a special type of
boundary conditions was applied at the wall/fluid interface that takes
into account a finite wall conductivity, as proposed in Walker (1981). It
should be noted that this approach can be applied only for a very thin
wall thickness and small conductance ratios.
Fusion engineering and technology-related research include

* Corresponding author.
E-mail address: S.Kenjeres@tudelft.nl (S. Kenjereš).

https://doi.org/10.1016/j.ijheatfluidflow.2021.108800
Received 16 August 2020; Received in revised form 1 February 2021; Accepted 22 February 2021
Available online 18 May 2021
0142-727X/© 2021 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

numerous topics dealing with the MHD phenomena. Smolentsev et al. (constant) magnetic field through the Lorentz force. Conservation of
(2015) provided an extensive review of MHD codes for fusion applica­ mass and momentum are used to describe the MHD flow (under the
tions and selected benchmark problems of importance for fusion appli­ assumption that the imposed magnetic field is known), and are written
cations. The proposed benchmarks covered a series of 2D and 3D steady as:
and developing MHD flows in both laminar and turbulent regimes, and
∇⋅U = 0 (1)
the final case also included effects of thermal buoyancy. Gajbhiye et al.
(2018) validated their general-purpose solver by analyzing the free ∂U 1 1
convection in a cubical enclosure under a uniform magnetic field and the + (U⋅∇)U = − ∇p + ν∇2 U + (J × B) (2)
∂t ρ ρ
electro-magnetically driven flow in a toroidal duct. The commercial
ANSYS-CFX finite-volume based code was used to simulate a water- where U is velocity, p is pressure, ν is the kinematic viscosity, ρ is den­
cooled lithium lead (WCLL) breeding blanket module subjected to a sity, J is the current density and B is the imposed magnetic field. In the
strong uniform magnetic field, Tassone et al. (2017). The commercial momentum equation, the MHD interactions are accounted for through
multi-physics finite-element code COMSOL was successfully applied to the Lorentz force term. In addition to the velocity and pressure, also the
simulate transient natural convection phenomena under influence of the current density (J) needs to be calculated. For the one-way coupled
imposed uniform magnetic field, Sahu and Bhattacharyay (2018). MHD phenomena, i.e. when the following conditions are valid
Validation of the multi-phase MHD flows is a challenging topic. The
number of validation studies dealing with multi-phase MHD phenomena UL ν
Rem = ≪1 and Prm = ≪1 (3)
is significantly smaller compared to single-phase MHD phenomena. The λ λ
analytical solutions for the multi-phase MHD situations are very scarce.
where Rem is the magnetic Reynolds number, Prm is the magnetic Prandtl
One of the recently proposed analytical solutions for a 2D multi-phase
number, L is the characteristic length and λ is the magnetic diffusion, ν is
MHD flow is presented in Righolt et al. (2016), where the elevation of
the kinematic viscosity, the Ohm’s law for a moving conducting fluid is
the liquid–metal/air interface due to the presence of an imposed mag­
used
netic field is analytically solved. Numerical simulations of a rising
bubble in the liquid metal subjected to an external homogeneous mag­ J = σ( − ∇ϕ + U × B) (4)
netic field of different strengths were studied in Shibasaki et al. (2010).
The finite-difference code was used and the terminal bubble velocity where σ is the electrical conductivity of the fluid. By imposing the
dependency on the strength of the imposed magnetic field was analyzed. divergence-free current density condition in the Ohm’s law, i.e.
Finally, the turbulent MHD phenomena require a special solving ∇⋅J = 0 (5)
strategy due to the necessity to properly capture both – the flow and
electromagnetic instabilities. The presence of the fluctuating Lorentz the final Poisson’s equation for the electric potential (ϕ) is obtained and
force requires a proper adaptation of the RANS-type of turbulence can be written as
models (Kenjereš and Hanjalić, 2000; Kenjereš et al., 2004) or applica­ ( )
tions of the eddy-resolving simulation techniques such as Direct Nu­ ∇2 ϕ = ∇⋅ U × B (6)
merical Simulations (DNS) or Large Eddy Simulations (LES), Kenjereš In addition to Rem and Prm (given in Eqn. (3)), the hydrodynamic
(2018). Krasnov et al. (2008) compared different sub-grid scale models Reynolds and Hartmann number are used as typical MHD non-
for the MHD LES channel flow and demonstrated ability of the dynamic dimensional parameters:
Smagorinsky model to properly predict the influence of the imposed √̅̅̅̅̅
magnetic field. Chaudhary et al. (2010) used DNS and analyzed how the Re =
UL
, Ha = BL
σ
(7)
increasing strength of a transverse magnetic field could influence the ν ρν
turbulence in the square duct flow. Mao et al. (2017) simulated the MHD
flow in the insulated squared duct with different sub-grid scale models 2.2. Governing equations for a multi-phase MHD: volume of fluid method
and compared data with the DNS results from the previous research of
Chaudhary et al. (2010). Additionally, Mao et al. (2017) varied the In the current study, the Volume of Fluid (VOF) method is applied to
Hartman number, showing how the turbulence is being suppressed by the multi-phase MHD flow simulations. In addition to the Lorentz force,
the imposed magnetic field. also the surface-tension and gravitational forces need to be included into
The main goal of the present study is to obtain and validate results the momentum equation:
from our newly developed OpenFOAM solver over a range of various
∂U 1
magnetohydrodynamic flows, and based on these findings, to propose an + (U⋅∇)U = − ∇p + νav ∇2 U
∂t ρav
extensive numerical MHD benchmark, which can be potentially useful (8)
1 ( )
for developers of the computer codes for simulations of the MHD phe­ + (J × B) + f g + γk∇α
ρav
nomena. We are primarily focusing on the influence of the finite electric
conductivity of surrounding walls and the multiphase aspects of the
where f g is the gravity force term, γ is the surface tension, k is the cur­
MHD phenomena. We have analyzed the following situations: (i) a
vature of the interface (calculated as k = ∇⋅|∇ ∇α
α|), νav is the phase aver­
laminar duct flow with finite conductivity of surrounding walls, (ii) a
laminar back-step flow, (iii) a shallow 2D multi-phase cavity, (iv) a aged viscosity (calculated as νav = α⋅ν1 + (1 − α)⋅ν2 , where ’1’ and ’2’
rising bubble in the liquid metal, and, finally, (v) a turbulent duct flow are phase indicators), ρav is the phase averaged density (calculated as
with conducting walls. For all mentioned cases we performed a detailed ρav = α⋅ρ1 + (1 − α)⋅ρ2 ) and the volume fraction α is described by the
comparative assessment against available analytical solutions or/and following transport equation:
numerical results presented in the literature. ∂α
( )
+ ∇⋅ αU + ∇⋅(Ur α(1 − α)) = 0 (9)
∂t
2. Governing equations and numerical details
where Ur is the artificial compression velocity used for the interface
2.1. Governing equations for a single-phase MHD sharpening, which is calculated as:
[ ( )]
|ψ | |ψ |
We consider an incompressible electrically conductive fluid with Ur = nf min Cα ⃒⃒ ⃒⃒, max ⃒⃒ ⃒⃒ (10)
liquid metal properties. The fluid is affected by the imposed external Sf Sf

2
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

where nf is the normal vector of the cell surface, ψ is the mass flux
through the face, Sf is the cell surface area, and Cα is a coefficient that is
used to control the interface thickness. There is no artificial interface
compression when Cα = 0. In order to control the spurious velocities
which appear near the interface due to the sharp change of α, the volume
fraction function is smoothed by the following Laplacian filter (Hoang
et al., 2013; Mukherjee et al., 2018): Fig. 1. Sketch of the fluid/wall interface condition for the conjugate
MHD problem.
∑
n
(αf Sf )
(11)
f =1
α̃c = ∑n ψ mhd = σf (U × B)f ⋅Sf (17)
(Sf )
f =1
where the cell -face electric conductivity (σ f ) is calculated by
where α ̃ is the resulting smooth volume fraction function, while sub­ applying the harmonic average between different phases, and (Sf ) is
scripts c and f indicate the cell center and cell face, respectively. Using the cell surface area vector.
the smooth function ̃ α in Eqn. (9), instead of the original function α will 2. Use Eq. (17) to solve the discretized electrical potential equation and
suppress these parasitic velocities. In the current study, the filter (11) is find electric potential (ϕ) at the cell centers:
applied twice for each time step. ∑
m ∑
m
σf ∇nf ϕ⋅|Sf | = ψ mhd (18)
f =1 f =1
2.3. The eddy-resolving MHD turbulence: large Eddy simulation
where ‘m’ indicates the number of cell faces.
Turbulence modeling is performed by the Large Eddy Simulation 3. Calculate the current density flux at cell faces using the surface-
(LES) method which is a good alternative to a more computationally normal gradient of electric potential (ϕ):
expensive Direct Numerical Simulations (DNS). We apply the spatial ⃒ ⃒
Jn = − σ f ∇nf ϕ⋅⃒Sf ⃒ + ψ mhd (19)
filtering operation to Eqn. (2), which finally can be written as:

∂U 1 1 where (Jn ) is the cell face normal component of the current density.
+ (U⋅∇)U = − ∇p* + ν∇2 U − ∇⋅τsgs + (J × B) (12) 4. Finally, use the current density flux from Eq. (19) and calculate the
∂t ρ ρ
fully conservative form of the Lorentz force as:
where (’’) indicates the spatially filtered value and (τsgs ) is the sub-grid
1 ∑ m

scale stress tensor, and p* = p + 13 τ′ I. The eddy viscosity concept is (J × B)c = −

Ωc f =1
(Jn )f (Bf × rf )
applied for the closure of the subgrid stress tensor as: (20)
1 ∑ m
( )12 − rc × (Jn )f Bf
Ωc f =1
τsgs
ij = − 2νsgs Sij , νsgs = (CS Δ)2 S, S = 2Sij Sij (13)

where (Ωc ) is the volume of cell, (rc ) is the cell center distance vector
where S is the modulus of the strain rate tensor, νsgs is the subgrid scale
and (rf ) is the face center distance vector.
turbulent viscosity, and Cs is Smagorinsky coefficient. In the present
work, we have adopted the dynamic approach to locally estimate values
2.4.2. Conjugate MHD: taking into account electric conductivity and
of the Smagorinsky coefficient, Lilly (1992), as follows:
thickness of surrounding walls
〈 〉
1 Lij Mij The finite electric conductivity and finite thickness of surrounding
C2S = 〈 〉 (14) walls have a significant impact on the fluid flow. This is due to the effects
2 Mij Mij
of the current density transfer between a liquid layer and solid walls,
which is directly influencing the intensity and direction of the local
Lij = Ũ ̃ ̃
i Uj + Ui Uj (15)
Lorentz force in the near-wall region. To include the fluid/wall interface
effects, we have developed an approach similar to traditional conjugate
Mij = Δ2 S̃ ̃2 ̃
Sij + Δ S̃Sij (16) heat transfer, but now instead of the heat flux transfer, we focus on the
distribution of the electric potential and current density in both do­
where Δ is the first filter (calculated as Δ = (Δx Δy Δz )1/3 ), Δ
̃ is the mains. Transport equations of the electric potential in liquid (L) and
second filter (calculated as Δ = 2Δ) and ’〈…〉’ means the local spatial
̃ solid (S) wall domains can be written as:
averaging over the cell faces.
∇⋅(σL ∇ϕL ) = ∇⋅(σL (U × B)) (21)

∇⋅(σS ∇ϕS ) = 0 (22)

2.4. Numerical details
Note that the source term (the RHS of Eqn.(21) is absent for the solid
2.4.1. The conservative form of the Lorentz force wall domain. Along the fluid/wall interface (Fig. 1), the conservation
The additional Lorentz force in the momentum equation is tradi­ and continuity of the electric current density (J) needs to be kept. This is
tionally treated in a non-conservative way (i.e. by applying the volume achieved by imposing following set of the boundary conditions at the
integration of the source term). This can potentially lead to significant interface:
numerical errors, especially for flow regimes with high Hartmann
∂ϕL ∂ϕ
numbers. Similarly, the total electric current density must be conserved σL = σS S (23)
too. Both of these requirements are achieved through the application of ∂n ∂n
the Four Steps Projection Method (FSPM) proposed by Ni et al. (2007), ϕL = ϕS (24)
which can be summarized through the following four steps:
The electric current density in the computational cell center is
1. Calculate the magnetic flux at cell faces: calculated in the same manner for both liquid and solid part of

3
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

Re = 10 and Hartmann number is varied in the 0⩽Ha⩽104 range. At the

inlet, a uniform velocity profile is imposed. At all walls, the no-slip ve­
locity boundary conditions are applied. At the outlet, a zero-pressure
boundary condition is imposed. The uniform transverse magnetic field
is imposed. To deal with the finite-thickness surrounding walls, we
introduce characteristic wall conductance parameter, defined as:
Cd = (σS ds )/(σ L L). (26)
Three types of electric boundary conditions for the walls perpen­
dicular to the magnetic field (Hartmann walls) are considered: (i)
Fig. 2. Sketch of the simulated domain for a laminar MHD flow in a duct with arbitrary conductive walls with varied wall conductance parameter
Hartmann walls with finite electric conductivity, subjected to a transverse (0.005⩽Cd ⩽40), (ii) fully electrically insulated walls (∂ϕ/∂n = 0 and
magnetic field. dS = 0), and finally, (iii) fully conductive walls (ϕ = 0 and dS = 0). The
walls parallel to the magnetic field (Shercliff walls) are considered as
computational domain as: electrically insulated for all cases.
m ( ) Although the final steady-state results are validated against analyt­
1 ∑ ( ) 1 ∑m
( )
Jc[L,S] = Jn,f [L,S] ⋅rf − rc ⋅ Jn,f [L,S] (25) ical solutions, the solution procedure is performed in a time-dependent
Ωc f =1 Ω f =1 mode. This time-dependent approach is not numerically efficient, but
our final goal is to have a well-validated solver able to simulate MHD
where the harmonic average is used to interpolate the electric conduc­ phenomena in transient and turbulent flow regimes, so we adopted a
tivity at the interface, needed for calculation of the current density flux time-dependent solution approach for all benchmark cases presented
at the cell faces (Jn,f ). here. The second-order central difference scheme (CDS) is applied for
both convective and diffusive terms of discretised momentum equation,
2.4.3. The computer code whereas the second-order backward scheme is used for time integration.
The integrated MHD solver, which includes all above-listed transport For all simulations the same hexahedral non-uniform orthogonal
equations, for both single- and multi-phase MHD phenomena is based on mesh is used with (Nx× Ny × Nz = 80 × 100 × 100)fluid control volumes
the finite-volume open-source computer code OpenFOAM-extend 4.0, for the fluid domain and (Nx × Ny × Nz = 80 × 10 × 100)solid for the
Weller et al. (1998). Coupling between pressure and velocity field is solid domain, respectively. In making the spatial distribution of the non-
performed with the PISO algorithm, Issa et al. (1986). uniform mesh, special attention is devoted that characteristic Hartmann
and Shercliff boundary layers (with a typical thickness of δHa = L/Ha
3. Applications: test cases
and δSh = L/Ha1/2 ) are properly resolved. This is achieved by placing
between 5 and 10 control volumes with a typical grid expansion ratio of
3.1. Laminar duct flow with conjugate MHD
1.14 in the region bounded by the wall and the edge of the boundary
layer (at δHa ).
In the first test case, we address a laminar pressure-driven flow of an
Contours of the calculated streamwise velocity and electric potential-
electrically conducting fluid in the rectangular duct subjected to a
after reaching steady state in the center of the duct (x = 10L) – are
transverse magnetic field, Fig. 2. The duct has the square cross-section
shown in Fig. 3. For the MHD neutral case (Ha = 0) the velocity exhibits
(where L- is the half-width), length of 20L and ds is the thickness of
a typical symmetric parabolic-like distribution, Fig. 3(a). By imposing
side-walls (Hartmann walls). The Reynolds number is kept constant at
the transverse magnetic field (Ha = 100) and by keeping all duct-walls

Fig. 3. The contours of the streamwise velocity (top row) and electric potential with current density streamlines (bottom row) in the center of the conjugate MHD
duct flow: (a), (e) Ha = 0. (b), (f) Ha = 100, fully insulated walls (Cd = 0). (c), (g) Ha = 100, arbitrary conductive walls (Cd = 0.1). (d), (h) Ha = 100, fully
conductive walls (Cd →∞).

4
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

Fig. 4. The streamwise velocity profiles along y-axis (between Hartmann walls) and z-axis (between Shercliff walls) in the duct at various Cd and Ha.

electrically insulated, a flattening of the velocity distribution occurs in

Fig. 5. Numerical mesh dependency on the streamwise velocity profiles along
z-axis in a conjugate MHD duct, Cd = 0.5, Ha = 50.
the central part of the duct, whereas thin Hartmann boundary layers are
generated along opposite vertical walls, Fig. 3(b). Next, by keeping the
same strength and direction of the imposed magnetic field, and by
changing electric properties of the vertical walls from fully insulated to
walls with a finite thickness and conductivity (i.e. Cd = 0.1), we observe
a dramatic reorganization of the velocity with peaks in the proximity of
the Shercliff walls, Fig. 3(c). Finally, by making Hartmann walls fully
electrically conducting (Cd →∞), the velocity distribution with two
peaks is still present, Fig. 3(d). The electric potential contours exhibit
close to a linear distribution in the vertical direction for electrically
insulated and finite-conductivity Hartmann walls, Fig. 3(f) and (g). In
contrast to this, the perfectly electrically conducting Hartmann walls
impose almost a uniform distribution in the central part of duct, Fig. 3
(h).
The numerical solutions are compared next against the following
analytical solutions: (1) Shercliff’s solution for the electrically insulated

Fig. 6. The streamwise velocity profiles along z-axis (between Shercliff walls) in the duct with arbitrary conductivity walls, (Cd = 0.05) for Ha = 5 × 103 (a) and
Ha = 104 (b), respectively.

5
A. Blishchik et al.
Fig. 7. The sketch of the simulation domain of the 2D laminar MHD back-step
test case.
walls, Shercliff (1953), (2) Hunt’s solution for the electrically fully
conductive walls, Hunt (1965), and (3) Sloan’s solution for the walls
with the arbitrary electrical conductivity and thickness, Sloan and Smith
(1966). For all simulated cases, an excellent agreement between present
numerical simulations and analytical solutions is obtained, confirming
an adequate implementation and validation of the conjugate MHD
solver, Fig. 4.
To illustrate the sensitivity of the numerical solution, we perform a
mesh dependency study with three mesh levels: (i) the coarse mesh
(M1), (Nx × Ny × Nz = 40 × 50 × 50)fluid and (Nxs × Nys × Nzs = 40×
5 × 50)solid = (0.11 × 106 )total CVs, (ii) the present mesh (M2)
(Nx × Ny × Nz = 80 × 100 × 100)fluid and (Nxs × Nys × Nzs = 80 × 10
×100)solid = (0.88 × 106 )total CVs and (iii) the fine mesh (M3)
(Nx × Ny × Nz = 160 × 200 × 200)fluid and (Nxs × Nys × Nzs = 160 ×
20 × 200)solid = (7.04× 106 )total CVs. As it can be seen in Fig. 5, a very
good agreement with the analytical solution is obtained for the inter­
mediate (M2) and fine mesh (M3), and that a slight underprediction of
the double peaks is observed for the coarse mesh (M1). To test possible
limits of the numerical stability and accuracy, two additional high
values of Ha = 5000 and 10000 are simulated for the case with a finite
electrically conducting walls (Cd = 0.05), Fig. 6. For such high values of
Ha, very strong wall jets are generated along Shercliff walls. With a
proper mesh refinement in the proximity of walls, i.e.
(Nx × Ny × Nz = 100 × 180 × 180)fluid control volumes for the fluid
domain and (Nx × Ny × Nz = 100 × 15 × 180)solid for the solid domain,
Again, a very good agreement is obtained between numerical simula­
tions and analytical solutions for both values of Ha, additionally proving
accuracy and numerical stability of the algorithm.
3.2. The 2D MHD laminar back-step flow
Next, we consider the two-dimensional backward-facing step flow in
a laminar flow regime subjected to a uniform vertical magnetic field,
Fig. 7. In contrast to the previous case, this configuration is expected to
produce a more complex flow pattern with a well-defined recirculation
region in the lower part of the domain. The channel height is L and its
length is 15L. The lower and upper boundaries of the channel are no-slip
walls. The upper half of the left boundary is the inlet, while the lower
half is the solid wall. The inlet velocity is defined as:
Fig. 8. The contours of the non-dimensional horizontal (streamwise) velocity (u/u0 ) for Re = 800 and different Ha. (a) Ha = 0, (b) Ha = 5, (c) Ha = 10, (d) Ha =
50.
6
International Journal of Heat and Fluid Flow 90 (2021) 108800
{
12(y − 1)(1 − 2y), L/2 < y < L
u(x = 0, y) = (27)
0, 0 < y < L/2
For the right boundary, a simple zero-gradient condition is imposed.
All walls are treated as perfectly electrically insulated. The simulation
domain and all boundary conditions are selected such that they match
exactly the numerical study of Mramor et al. (2014), who applied a MHD
extension of the Local Radial Basis Function Collocation Method
(LRBFCM). The entire simulation domain is represented by an orthog­
)
onal numerical mesh with (Nx × Ny = 600 × 50) = (3 × 104 total control
volumes. Two values of the Reynolds number are simulated (Re = 300
and 800, where Re = ux L/ν) over a range of Hartmann numbers
(0⩽Ha⩽50). The second-order linear upwind differential scheme is used
for convective terms, the second-order central differencing scheme
(CDS) is used for diffusion terms, and the second-order backward
scheme for the time-integration. The contours of the non-dimensional
streamwise velocity (u/u0 where u0 = (ux )|x=0 , i.e. the inlet integrated
velocity profile), at Re = 800 and different strengths of the imposed
magnetic field (Ha = 0, 5, 10 and 50) are shown in Fig. 8. It can be seen
that with a magnetic field increase, the recirculation length reduces, and
flow becomes much more uniform. At Ha = 0, two large recirculation
regions along the upper and lower walls are generated. With Ha in­
crease, the recirculation region along the upper wall disappears, while
the recirculation long the lower wall is still present, but its length is
significantly reduced. This reduction of the recirculation region is
further illustrated in zoom-in plots, where we superimposed contours of
the streamwise velocity and streamlines, as shown in Fig. 9. At the
highest value of Ha = 50, the recirculation can be observed only in a
very small region attached to the lower part of the inlet plane. A
Fig. 9. Same as in the previous figure, only now the zoom-in regions in the
proximity of the inlet are shown with superimposed streamlines.
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

Fig. 10. The vertical profiles of the non-dimensional horizontal (u/u0 ) and vertical (v/u0 ) velocity components at various Ha and two values of Re: Re = 300 (a–b)
and Re = 800 (c–d). Comparison between the reference study based on the Local Radial Basis Function Collocation Method (LRBFCM) (Mramor et al., 2014) and the
present Finite Volume Method (FVM) results.

Fig. 11. The non-dimensional horizontal (u/u0 ) and vertical (v/u0 ) velocity profiles at the exit plane for various meshes at Re = 800 and Ha = 5 compared to the
reference solution (LRBFCM, Mramor et al., 2014).

comparison of obtained profiles of horizontal (u/u0 ) and vertical (v/u0 )

velocity components at the exit plane with values presented in the
literature (Mramor et al., 2014), are shown in Fig. 10. It can be seen that
the horizontal velocity profiles become more flat with the magnetic field
increase for both Reynolds numbers. The vertical velocity component
almost completely disappears at higher values of Ha. A very good
agreement between the present profiles and results from the literature
(Mramor et al., 2014) is obtained for all presented cases. To demonstrate
that the obtained results at present mesh of (Nx × Ny = 600 × 50) (M2)
(3 × 104 CVs) are grid independent, one coarser (Nx × Ny = 300 × 25)
Fig. 12. The sketch of the simulation domain of a two-dimensional multi-phase
(M1) (0.75 × 104 CVs) and one finner (Nx × Ny = 1200 × 100) (M3) MHD cavity test case.
(1.2 × 105 CVs) numerical mesh are generated, and results are compared
in Fig. 11. A good agreement between different mesh levels is obtained, netic field and electric potential difference. The two-dimensional cavity
with a slight overprediction of the local maxima of the non-dimensional with characteristic length L and partially filled with the electrically
vertical velocity (v/u0 ) at y/L = 0.7 for the coarse mesh. conductive liquid (where d is the liquid layer height and d≪L) is shown
in Fig. 12. The upper part of the cavity if filled with air (σ air = O (10− 15 )
3.3. The multi-phase two-dimensional shallow cavity flow with MHD S/m, i.e. negligible electric conductivity). The external magnetic field is
aligned with the negative z-direction (perpendicular to the cavity) and
The first example of the MHD multi-phase test case is a shallow its linear distribution is defined as:
cavity subjected to combined effects of the imposed non-uniform mag­

7
A. Blishchik et al.
Fig. 13. The velocity vector distribution (a) and contours of the non-
dimensional horizontal (x-component) velocity (Ux /U* ) in the 2D MHD cav­
ity, Re* = A, Ha* = 1, Bo* = A2 , Ca* = A4 .
B = − b0 (1 + αb ⋅x)̂
z (28)
where αb = 0.1 defines a distribution parameter. The no-slip velocity
boundary condition is imposed at all walls (i.e. bottom and side-walls).
The gravity force is aligned with the negative y-coordinate direction.
The side-walls are kept at constant (but different) electric potential
(ϕ1 = − 12 Δϕ, ϕ2 = 12 Δϕ, where Δϕ is the imposed electric potential
difference). The bottom wall is perfectly electrically insulated (∂ϕ/∂n =
0 and Cd = 0). Because of the imposed magnetic field and electric po­
tential difference, the generated Lorentz force within the fluid will drive
the flow. This fluid motion will be opposed by a joint combination of the
viscous, gravitational, and surface tension forces. To account for addi­
tional free-surface related physical mechanisms, the following set of
non-dimensional parameters is introduced, Righolt et al. (2016):
√̅̅̅̅̅̅̅̅
U*d σαb ρgd2 ρνU *
Re* = , Ha* = b0 d , Bo* = , Ca* = (29)
ν μ γ γ
In addition to the redefined Reynolds (Re* ) and Hartmann (Ha* ), also
the Bond (Bo* ) and capillary (Ca* ) numbers are introduced. The char­
acteristic non-dimensional velocity is calculated as:
σ Δϕb0 αb dA
U* = (30)
ρν
Because of the large number of possibilities based on the various
combinations of characteristic non-dimensional numbers, in the present
work we kept constant Re* = A and Ha* = 1, while we change Bo* and
Ca* . We also kept the identical aspect ratio of the domain, A = d/L =
0.1. The two-dimensional orthogonal, non-uniform mesh (Nx × Ny =
Fig. 14. The free-surface elevation for various Ca and Bo. Comparison between the present simulations (CFD) and analytical solution of Righolt et al. (2016).
8
International Journal of Heat and Fluid Flow 90 (2021) 108800
50 × 200) with rectangular control volumes is used. The central differ­
encing scheme (CDS) is used for the diffusive and convective terms of
transport equations. The time-integration is performed with the second-
order backward scheme. For this particular case, the different values of
the interface compression coefficient (0⩽Cα ⩽1) did not have any sig­
nificant impact on the obtained solutions due to a smooth free-surface
deformation. The local variation of the resulting Lorentz force gener­
ates the flow of electrically conducting fluid (initially at rest) in the
lower part of the cavity with characteristic elevation of the free-surface,
as shown in Fig. 13.
This non-dimensional vertical elevation (h/d) of the free-surface, as a
function of Ca* and Bo* numbers, is shown in Fig. 14. It can be seen that
an excellent match between the present numerical results (CFD) and
analytical solutions is obtained for all calculated cases. Note that a
vertical elevation of the free-surface increases with an increase in both
Ca* and Bo* . The horizontal profiles of the non-dimensional horizontal
(u/u* ) and vertical (v/u* ) velocity in the proximity of the left-wall are
shown in Fig. 15(a) and (b), respectively. The vertical profile of the non-
dimensional horizontal velocity at the central vertical line is shown in
Fig. 15(c). Again, an excellent agreement between the present simula­
tion (CFD) and analytical solution from the literature (Righolt et al.,
2016) is obtained, proving the capability of the MHD multi-phase solver.
To confirm that the presented solutions are grid independent, we
analyzed the non-dimensional free-surface elevation (h/d) for three
mesh sizes: (i) the coarse mesh (M1) (Nx × Ny = 25 × 100), (ii) the in­
termediate (previously presented results) mesh (M2) (Nx × Ny = 50 ×
200), and (iii) the fine mesh (M3) with (Nx × Ny = 100 × 400). A good
agreement between results at different mesh resolutions confirms the
full mesh convergence of the presented results, Fig. 16.
3.4. The 3D rising gas bubble in liquid metal subjected to a longitudinal
magnetic field
A rising gas bubble (with an initial diameter db = L/2) is submerged
into the liquid metal confined in the 3D rectangular box (with height 3L,
width and depth L) is analyzed next, Fig. 17. This test case is based on
the study of Shibasaki et al. (2010). All boundary surfaces are electri­
cally insulated walls (∂ϕ/∂n = 0,Cd = 0) with imposed no-slip boundary
conditions. The external magnetic field is aligned with the y-coordinate
and the gravity is oriented in the opposite direction. The problem is fully
defined with the following set of non-dimensional parameters:
gρ2 d3 γρ d
√̅̅̅̅̅
σL
G = G2 b , Γ = G2 , Ha = Bdb (31)
μG μG μL
where G is the Galilei number, Γ is the Tension number, and subscripts
(G) and (L) indicate the gas and liquid phase, respectively. The non-
dimensional velocity, pressure, and time are defined as:
/ /( ) /
u* = μG (ρG db ), p* = μ2G ρG d2b , t* = ρG d2 μG (32)
We kept constant G = 4⋅104 , Γ = 2⋅106 and varied 0⩽Ha⩽200 to
A. Blishchik et al.
Fig. 15. The profiles of the velocity components in the proximity of the side-wall extracted along the y = d/2 line (a), (b), and in the center of the cavity extracted
along the x = 0 line (c): Re* = A, Ha* = 1, Bo = A2 , Ca = A4 .
Fig. 16. The mesh-dependency of the non-dimensional free-surface elevation
(h/d) for Re* = A, Ha* = 1, Bo = A2 , Ca = A4 .
study the influence of the magnetic field strength on the rising bubble
behavior. The electrical conductivity ratio is σ G /σ L = 2.49⋅10− 7 . The
orthogonal mesh is created with the mesh size (Nx × Ny × Nz = 60 ×
180 × 60), identical to the mesh used in Shibasaki et al. (2010). The
second-order linear-upwind scheme is used for the convective terms in
both momentum and volume fraction equations, whereas the backward
scheme is used for time integration. Because of a sharp jump of the
electrical properties at the phase interface, we have applied the har­
monic interpolation scheme for the electric conductivity. For this case,
the interface compression coefficient (Cα ) had stronger effect on the final
shape of the rising bubble. The selected value of Cα = 0.1 proved to be a
good choice for both multi-phase benchmarks presented here. The ob­
tained characteristic bubble shape, current density streamlines, contours
of the vertical velocity and pressure in the central vertical plane at an
arbitrary time instant t/t * = 0.02 and for Ha = 50, are shown in Fig. 18.
The current density streamlines form close loops around the bubble with
9
International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 17. The sketch of the simulation domain for the rising bubble in a liquid
metal subjected to an external (axial) magnetic field.
opposite directions above and below the bubble. The velocity contours
portray an updraft region in the center of the domain – above and below
the bubble, whereas the down-drafts are generated along the side walls.
Contours of the pressure exhibit almost linear distribution in the vertical
direction, with small deviations in the proximity of the bubble surface. It
can be seen that the resulting shape of the bubble strongly depends on
the imposed magnetic field strength, Fig. 19. The higher Ha leads to the
bubble stretching in the direction of the imposed magnetic field (y-di­
rection) and to a reduction of its rising velocity. We compare our results
with a numerical study of Shibasaki et al. (2010) who applied the finite-
difference (FDM) multi-phase MHD code. Comparison of the computed
terminal velocity for different values of Ha is shown in Fig. 20. After an
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

Fig. 18. The bubble shape (extracted as the isosurface of the volume fraction α = 0.5) with superimposed streamlines of the total current density (a), (b) contours of
the non-dimensional vertical velocity (uy /u* ) in the central vertical plane, (c) contours of the non-dimensional pressure field (p/p* ) in the central vertical plane - all at
t/t * = 0.02 and for Ha = 50.

Fig. 19. The bubble shape (identified as the isosurface of the volume fraction α = 0.5) and its location at time instant t/t * = 0.025 for various Ha: Ha = 0, 50, 100,
200 (a–d), respectively.

Table 1
The reattachment position (at y/L = 0 for Re = 300 and 800,
and 0⩽Ha⩽100).
Present LRBFCM, Mramor et al. (2014)
Re Ha x/L x/L

300 0 3.57 3.57

5 2.56 2.55
10 1.28 1.28
50 0.02 0.02
100 0.007 0.01

800 0 6.07 6.1

5 5.46 5.48
10 2.93 2.93
50 0.07 0.07
100 0.01 0.01
Fig. 20. Terminal bubble velocity at various Ha. Comparison with the finite-
difference (FDM) results of Shibasaki et al. (2010).

10
A. Blishchik et al.
Table 2
The non-dimensional terminal velocity at Ha = 50 and Ha = 200
for different meshes. Comparisons with values presented in the
finite-differences based method (FDM) of Shibasaki et al. (2010).
Ha The non-dimensional Present Shibasaki et al.
terminal velocity, (uy /u* ) (2010)
50 Mesh (M1) 143.3
Mesh (M2) 147.8 152
Mesh (M3) 149.9
200 Mesh (M1) 41.3
Mesh (M2) 42.1 58
Mesh (M3) 42.5
Fig. 21. Sketch of the simulation domain for the fully-developed (periodic flow
in the x-direction) turbulent MHD duct flow with Hartmann walls with finite
thickness (ds ) and electric conductivity (σ S ).
initial slight increase in the terminal velocity for intermediate values of
Ha < 50, a gradual decrease is obtained with a further increase of the
imposed magnetic field. The agreement between the current simulations
and data presented in Shibasaki et al. (2010) is good up to Ha = 50.
After reaching this peak value, larger differences are observed, but
qualitatively similar trends are observed. Differences for larger values of
Ha number can be partially explained by the use of different dis­
cretization approaches (the present finite-volume vs. finite-difference of
Shibasaki et al. (2010)), the application of different convective schemes
(the present second-order linear-upwind vs. the third-order UTOPIA
scheme of Shibasaki et al. (2010)), as well as due to the absence of the
mesh-dependency study of Shibasaki et al. (2010)). We also performed
additional simulations with a second-order quadratic-upwind scheme
for convective terms in momentum equations, but this resulted in mar­
ginal differences of rising velocity (less than 1%) compared to the linear-
upwind scheme (see Table 1).
Finally, we complete a mesh-dependency study for two different
Hartmann numbers Ha =50 and 200, and three meshes: (i) the coarse
mesh (M1) (Nx × Ny × Nz = 30 × 90 × 30) = (0.081 × 106 )total CVs, (ii)
the present mesh (M2) (Nx × Ny × Nz = 60 × 180 × 60) =
(0.64 × 106 )total CVs and (iii) the fine mesh (M3) (Nx × Ny × Nz = 120 ×
360 × 120) = (5.1 × 106 )total CVs. Results in Table 2 demonstrate that
the finest mesh (M3) provides the best agreement with the reference
data. However, the difference in terminal velocity values between in­
termediate (M2) and fine (M3) mesh is only 1%, while the total number
of CVs is four times larger. Based on this small difference, we conclude
that results are grid independent already at the mesh (M2).
3.5. A conjugate MHD duct flow in a fully developed turbulent regime
The final test case is a conjugate MHD square duct flow in a fully
developed turbulent regime. The duct height and width are L, and its
length is 16L. The imposed magnetic field is aligned with the y-axis and
perpendicular to the flow direction, Fig. 21. The periodic boundary
conditions are imposed in the streamwise (x-coordinate) direction. All
11
International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 22. The instantaneous coherent structures identified with the second-
invariant of the velocity-gradient tensor (Q-criterion) and colored by the
streamwise velocity, Re = 5602 (a) Ha = 0, (b) Ha = 21.2, insulated walls
(Cd = 0), (c) Ha = 21.2, arbitrary conductive walls (Cd = 0.05). The iso-
surface value Q = 1 s− − 2. Note that scaling factor of 2× is applied in the x-
direction to provide a more compact view.
other surfaces are walls with imposed no-slip velocity boundary condi­
tions. The lower and upper walls (Shercliff walls) are fully electrically
insulated (∂ϕ/∂n = 0 and Cd = 0). The front and back walls (perpen­
dicular to the imposed magnetic field – Hartmann walls) are considered
to have three different types of electric boundary conditions: (i) the
finite conducting walls with the wall conductance parameter Cd =
(σ S dS )/(σL L) = 0.05, (ii) fully electrically insulated walls, and (iii) fully
conductive walls (ϕ = 0 and Cd →∞). We apply the wall-resolving dy­
namic large-eddy simulation (LES) approach. The numerical mesh
)
contains (Nx × Ny × Nz = 240 × 120 × 120 fluid = (3.456 × 106 ) CVs
)
and (Nx × Ny × Nz = 240 × 12 × 120 wall = (0.3456 × 106 ) CVs in the
fluid and wall regions, respectively. The non-dimensional mesh pa­
rameters are Δy+ wall = Δzwall ≈ 0.6, Δycore = Δzcore ≈ 6 and Δx ≈ 25.
+ + + +
The central differencing scheme (CDS) is used for spatial discretization
and the second-order backward scheme for temporal discretization. The
flow is defined with the following set of the non-dimensional parame­
ters: Re = 5602 and Ha = 21.2. Furthermore, a simulation with Ha = 0
is performed in order to provide a comparison with the non-MHD
neutral case. The selected value of the Reynolds number assures that a
fully developed turbulence is generated and maintained. All simulations
are statistically averaged over at least 100 flow-through times, and the
spatial averaging procedure is applied to accelerate the convergence of
the flow statistics. The instantaneous coherent structures colored by
streamwise velocity for various Ha, and Hartmann walls conductivities
(expressed through the wall-conductivity parameter, Cd ), are shown in
Fig. 22. Under the action of the imposed transverse magnetic field, by
changing the electric properties of the walls, the coherent structures
start to be suppressed in the proximity of Hartmann walls, as seen from
the side-views of the duct shown in Fig. 22(a)–(c).
Contours of the long-term time-averaged streamwise velocity, tur­
bulent kinetic energy, and electric potential, for various wall conduc­
tivities, are shown in Fig. 23. Starting from a symmetrical distribution,
contours of the mean streamwise velocity start to be suppressed in the
direction of the imposed magnetic field (y-direction). This behavior is
caused by the reorganization of the electric current density streamlines.
In contrast to the fully closed current loops in the fluid region (for
electrically insulated walls), a finite electric conductivity of walls makes
that current density loops also enter these regions, causing significant
changes in resulting Lorentz force components in the y- and z-directions,
respectively.
The contours of the turbulent kinetic energy portray the reorgani­
zation from fully symmetrical distributions for the neutral case with
characteristic peaks in the proximity of duct walls, Fig. 23(e), to the non-
symmetrical distributions for non-insulated walls, Fig. 23(f)–(h). It can
be seen that the levels of turbulent kinetic energy are suppressed in the
proximity of Hartmann walls for the fully insulated and walls with finite
conductivity, Fig. 23(f)–(g). At the same time, distributions of the
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800

Fig. 23. The long-term time-averaged contours of the streamwise velocity (top row), turbulent kinetic energy (middle row) and electric potential (bottom row) in the
central vertical plane of the duct shown in Fig. 21, for a fixed value of Re = 5602: (a–e–i) Ha = 0, (b–f–j) Ha = 21.2 with fully electrically insulated walls (Cd = 0),
(c–g–k) Ha = 21.2 with arbitrary conducting walls (Cd = 0.05), (d–h–l) Ha = 21.2 with fully conducting walls (Cd →∞).

Fig. 24. The long-term time-averaged non-dimensional streamwise velocity (the semi-log plots of U+ vs. y+ and z+ ) profiles (where U+ = U/Uτ ,y+ = yUτ /ν, and) in
the proximity of Hartmann (a) and Shercliff wall (b), respectively.

turbulent kinetic energy along Shercliff walls are just slightly affected by
changing Hartmann walls conductivity, Fig. 23(f)–(g). Interestingly, for
the fully conducting walls, the turbulence kinetic energy along Shercliff
walls is reduced in comparison to values along Hartmann walls, Fig. 23
(h).
The contours of the electric potential also illustrate significant
changes between the fully electrically insulated Fig. 23 and fully con­
ducting Hartmann walls Fig. 23(l), with the latter exhibiting signifi­
cantly more pronounced non-uniform distribution in the vertical
direction. Next, we move to a more detailed comparison of the charac­
teristic long-term time-averaged first- and second-order statistics
profiles, Figs. 24 and 25. The present results obtained with the dynamic
LES approach are validated against two reference studies: Gavrilakis
(1992) who simulated an MHD neutral turbulent duct flow, and
Chaudhary et al. (2010) who simulated MHD turbulent duct flow at
Ha = 21.2 with fully electrically insulated walls – both using the fully-
resolving Direct Numerical Simulations (DNS) approach. The time-
averaged mean streamwise velocity profiles in the proximity of Hart­
mann and Shercliff walls are shown in Fig. 24(a) and (b), respectively. It
can be seen that a very good agreement between the present and DNS
results from the literature is obtained at both locations. The profiles of
the non-dimensional rms values of the fluctuating streamwise velocity

12
A. Blishchik et al.
√̅̅̅̅̅̅̅
2
Fig. 25. The non-dimensional rms of streamwise velocity ( u′ /Uτ ) in the proximity of Hartmann (along y/L) (a) and Shercliff (along z/L) wall (b), respectively, for
different Hartmann wall conductivities.
along the identical locations reveal an interesting behavior, Fig. 25. In
the proximity of Shercliff walls, an increase of the wall conductivity
produced a gradual decrease of the rms values, Fig. 25(b). In contrast to
this behavior, the distributions in the proximity of Hartmann walls
indicate an initial suppression for the fully electrically insulated walls,
followed by an increase for the fully conducting walls, Fig. 25(a). Again,
a good agreement with available DNS references (for the non-MHD sit­
uation and the MHD case with fully insulated walls) is obtained con­
firming suitability of here used dynamic LES approach.
4. Summary and conclusion
We have presented a comprehensive numerical benchmark study
addressing a range of single- and multi-phase one-way coupled MHD
flows. The single-phase cases included the conjugate MHD flows in ducts
with varied electric conductivity of the wall – in both laminar and tur­
bulent flow regimes, and the laminar back-step flow subjected to a
transverse magnetic field. The multi-phase cases covered a two-
dimensional MHD cavity and a rising bubble in a liquid metal flows –
both simulated with the volume of fluid (VOF) approach. We have
implemented an extended set of MHD transport equations in the open-
source code OpenFOAM. Our particular focus was to extend the exist­
ing set of MHD benchmarks and to provide a detailed comparison with
similar studies in the literature. We also proposed a novel methodology
and benchmark for a conjugate MHD in a turbulent duct flow with an
arbitrary wall conductivity (expressed in terms of the wall conductance
parameter). For the multi-phase flows, we have introduced a recently
proposed analytical solution of a two-dimensional partially-filled cavity
flow subjected to an external magnetic field. An excellent agreement
was obtained for all cases for which analytical solutions are available.
For considered test cases without analytical solutions, a very good
agreement was obtained with available numerical studies from the
literature. It is concluded that here developed and validated version of
the computer code can be used for advanced fundamental and indus­
trial/technological studies involving various aspects of the MHD
phenomena.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgments
This project has received funding from the European Union’s Hori­
zon 2020 research and innovation program TOMOCON (Smart Tomo­
graphic Sensors for Advanced Industrial Process Control) under the
Marie Sklodowska-Curie grant agreement No. 764902.
13
International Journal of Heat and Fluid Flow 90 (2021) 108800
References
Hartmann, J., Lazarus, F., 1937. Theory of the laminar ow of an electrically conducting
liquid in an homogeneous magnetic field. Kongelige Danske Videnskabernes Selskab
Matematisk 15, 1–28.
Shercliff, J.A., 1953. Steady motion of conducting fluids in pipes under transverse
magnetic fields. Math. Proc. Cambridge Philos. Soc. 49, 136–144. https://doi.org/
10.1017/S0305004100028139.
Feng, J., Chen, H., He, Q., Ye, M., 2015. Further validation of liquid metal MHD code for
unstructured grid based on OpenFOAM. Fusion Eng. Design 100, 260–264. https://
doi.org/10.1016/j.fusengdes.2015.06.059.
Tao, Z., Ni, M., 2013. Benchmark solutions for MHD solver development. Sci. China Phys.
Mech. Astron. 56, 378–382. https://doi.org/10.1007/s11433-013-4997-5.
Walker, J., 1981. Magnetohydrodynamic flow in rectangular ducts with thin conducting
walls. J. Mecanique 20, 79–112.
Smolentsev, S., Badia, S., Bhattacharyay, R., Bühler, L., Chen, L., Huang, Q., Jin, H.-G.,
Krasnov, D., Lee, D.-W., Valls, E.M.D.L., et al., 2015. An approach to verification and
validation of MHD codes for fusion applications. Fusion Eng. Design 100, 65–72.
https://doi.org/10.1016/j.fusengdes.2014.04.049.
Gajbhiye, N.L., Throvagunta, P., Eswaran, V., 2018. Validation and verification of a
robust 3-D MHD code. Fusion Eng. Design 128, 7–22. https://doi.org/10.1016/j.
fusengdes.2018.01.017.
Tassone, A., Caruso, G., Nevo, A.D., Piazza, I.D., 2017. CFD simulation of the
magnetohydrodynamic flow inside the WCLL breeding blanket module. Fusion Eng.
Design 124, 705–709. https://doi.org/10.1016/j.fusengdes.2017.05.098.
Sahu, S., Bhattacharyay, R., 2018. Validation of COMSOL code for analyzing liquid metal
magnetohydrodynamic flow. Fusion Eng. Design 127, 151–159. https://doi.org/
10.1016/j.fusengdes.2018.01.009.
Righolt, B., Kenjereš, S., Kalter, R., Tummers, M., Kleijn, C., 2016. Analytical solutions of
one-way coupled magnetohydrodynamic free surface flow. Appl. Math. Model. 40
(4), 2577–2592. https://doi.org/10.1016/j.apm.2015.09.101.
Shibasaki, Y., Ueno, K., Tagawa, T., 2010. Computation of a rising bubble in an enclosure
filled with liquid metal under vertical magnetic fields. ISIJ Int. 50 (3), 363–370.
https://doi.org/10.2355/isijinternational.50.363.
Kenjereš, S., Hanjalić, K., 2000. On the implementation of effects of Lorentz force in
turbulence closure models. Int. J. Heat Fluid Flow 21 (3), 329–337. https://doi.org/
10.1016/s0142-727x(00)00017-5.
Kenjereš, S., Hanjalić, K., Bal, D., 2004. A direct-numerical-simulation-based second-
moment closure for turbulent magnetohydrodynamic flows. Phys. Fluids 16 (5),
1229–1241.
Kenjereš, S., 2018. On modeling and eddy-resolving simulations of flow, turbulence,
mixing and heat transfer of electrically conducting and magnetizing fluids: a review.
Int. J. Heat Fluid Flow 73, 270–297. https://doi.org/10.1016/j.
ijheatfluidflow.2018.09.003.
Krasnov, D., Zikanov, O., Schumacher, J., Boeck, T., 2008. Magnetohydrodynamic
turbulence in a channel with spanwise magnetic field. Phys. Fluids 20 (9), 095105.
https://doi.org/10.1063/1.2975988.
Chaudhary, R., Vanka, S.P., Thomas, B.G., 2010. Direct numerical simulations of
magnetic field effects on turbulent flow in a square duct. Phys. Fluids 22 (7),
075102. https://doi.org/10.1063/1.3456724.
Mao, J., Zhang, K., Liu, K., 2017. Comparative study of different subgrid-scale models for
large eddy simulations of magnetohydrodynamic turbulent duct flow in OpenFOAM.
Comput. Fluids 152, 195–203. https://doi.org/10.1016/j.compfluid.2017.04.024.
Hoang, D.A., Steijn, V.V., Portela, L.M., Kreutzer, M.T., Kleijn, C.R., 2013. Benchmark
numerical simulations of segmented two-phase flows in microchannels using the
volume of fluid method. Comput. Fluids 86, 28–36. https://doi.org/10.1016/j.
compfluid.2013.06.024.
Mukherjee, S., Zarghami, A., Haringa, C., van As, K., Kenjeres, S., van den Akker, H.,
2018. Simulating liquid droplets: a quantitative assessment of lattice Boltzmann and
Volume of Fluid methods. Int. J. Heat Fluid Flow 70, 59–78. https://doi.org/
10.1016/j.ijheatfluidflow.2017.12.001.
Lilly, D.K., 1992. A proposed modification of the Germano subgrid-scale closure method.
Phys. Fluids A Fluid Dyn. 4 (3), 633–635. https://doi.org/10.1063/1.858280.
A. Blishchik et al.
Ni, M.-J., Munipalli, R., Huang, P., Morley, N.B., Abdou, M.A., 2007. A current density
conservative scheme for incompressible mhd flows at a low magnetic Reynolds
number. Part II: on an arbitrary collocated mesh. J. Comput. Phys. 227 (1), 205–228.
https://doi.org/10.1016/j.jcp.2007.07.023.
Weller, H.G., Tabor, G., Jasak, H., Fureby, C., 1998. A tensorial approach to
computational continuum mechanics using object-oriented techniques. Comput.
Phys. 12 (6), 620. https://doi.org/10.1063/1.168744.
Issa, R., Gosman, A., Watkins, A., 1986. The computation of compressible and
incompressible recirculating flows by a non-iterative implicit scheme. J. Comput.
Phys. 62 (1), 66–82. https://doi.org/10.1016/0021-9991(86)90100-2.
Hunt, J.C.R., 1965. Magnetohydrodynamic flow in rectangular ducts. J. Fluid Mech. 21,
577–590. https://doi.org/10.1017/S0022112065000344.
14
International Journal of Heat and Fluid Flow 90 (2021) 108800
Sloan, D., Smith, P., 1966. Magnetohydrodynamic flow in a rectangular pipe between
conducting plates. Z. Angew. Math. Mech. 46, 439–443. https://doi.org/10.1002/
zamm.19660460705.
Mramor, K., Vertnik, R., Šarler, B., 2014. Simulation of laminar backward facing step
flow under magnetic field with explicit local radial basis function collocation
method. Eng. Anal. Boundary Elem. 49, 37–47. https://doi.org/10.1016/j.
enganabound.2014.04.013.
Gavrilakis, S., 1992. Numerical simulation of low-Reynolds-number turbulent flow
through a straight square duct. J. Fluid Mech. 244 (-1), 101. https://doi.org/
10.1017/s0022112092002982.